The diagram to the left shows a neural network training session over some 20,000 iterations. An iteration represents a training cycle, where the strength (or weight) of the neuron connections (see home page) are weighted according to the results of the previous cycle. The connections are given random weights to begin with.
The horizontal axis, which has a logarithmic scale, shows the number of iterations.
The blue graph shows the progress of learning by displaying the maximum example error for each iteration.
The red graph is the average error.
The orange graph is an indication of how well the network is validating.
Several different histograms can help to determine the health of the network and its capacity to train.
The histogram displayed is a section of the strength of the connection between input and hidden layers.
